Vessel

The rising tide lifts all boats And therefore we've rebranded to Vessel Same product and thesis but the name Vessel is more true to our mission: to build the most optimal craft to traverse the rising tide, no matter which model is cresting this week

@CWood_sdf @JFPuget 😂😂😂

i love how people are saying "if we write a sufficiently detailed specification, the agent can write all our code" do you know what writing a sufficiently detailed specification that deterministically maps to what a computer's actions is? it's coding

Great thread. What does this mean for OpenAI and Anthropic? Seems they’re doing the little guys a huge favor by lighting money on fire to build these sophisticated general purpose models that the smaller players can distill from for a fraction of the cost Obviously that’s great for advancement of the technology but seems silly from a competitive standpoint

Premium newsletter: The AI bubble is entirely circular, with 90% of revenues flowing through OpenAI and Anthropic. Outside of two companies, the majority of GPU customers need far fewer GPUs, meaning that the DeepSeek problem is industry-wide. wheresyoured.at/premium-ais-ci…

Who the hell uses Firefox?

If you've been laid off recently start a business it'll be the best thing that ever happened to you

What models are you all using in your products?

@staysaasy Yeah that’s helpful but not AGI

I don't think we're close to anyone vibe coding in-house versions of SaaS platforms with significant scale and complexity, so if you have a platform with a few meaty useful interconnected features you're probably pretty safe. But they're definitely gonna vibe code anything that is equivalent to: * Point an LLM at a Google Drive folder and ask it to do things that LLMs can already do * Search a really huge log file and do something SIMPLE with it at the end * Point an LLM at a spreadsheet and ask it questions And a reminder that if you squint hard enough a SQL database is shaped like a spreadsheet.

@asmartbear It’s all looksmaxxing

I’m so naive I thought “building in public” meant honestly sharing the ups and downs, so others can commiserate and encourage and celebrate and even give feedback. As opposed to posturing, going for “viral” posts, and sometimes just lying. Again, naive, I get it.

@atelicinvest And will these issues go away with ‘better’ models? Research says we’re plateauing and LLMs will never be AGI

You see, there's a reason why I'm skeptical as fuck about all this token usage. The current narrative goes something like this - Opus 4.5 was an inflection point - That + Claude Code = you can create anything by just telling it what to do - tokens go brrr - anthropic can't meet demand. they planned for super duper crazy growth and are getting super duppper suuper duuuper over 9000 growth - everyone is using tokens to do amazing, productive things. getting things done that would have taken a team of people months in an afternoon while sucking on a lollipop - the future is here. token demand is infinite. we are still so early. this is the leading edge. everyone is gonna work like this soon. What I'm seeing in the real world - guys spending 50m tokens to show you a dashboard of their wedding - guys spending 100m tokens to organize their email inbox - claude deleting production database in one shot - claude 4.7 ignoring direct instructions in a relatively short claude.md file every 5 minutes - no I don't have bunch of other shit installed. - If I work on more than 2-3 things at a time, I get slopfest- not because of code quality, but because the product direction becomes incoherent. - Every time I see someone producing billions of tokens, their main business is selling skills file, or being an influencer, or producing slopware bordering on fraudware or are doing masturbatory slopfork that nobody uses. - Real engineers are doing real work with it - 100% - but usage is nowhere near as high as you think it is - Guys that output actual quality software such as anomalyco telling you how difficult it is - Real software companies are actually having a lot of trouble adopting agentic coding tools beyond fooling around with PoC creation. - Whenever I audit personal tools created there are tons of hallucinated features or errors - the creators of which are blissfully unaware because none of them are anal retentive enough to go in there and check shit manually - no no the bot told me it was all good man! - these things are really good at giving the appearance of getting stuff done ... until you actually go into what it did and it's just slopping all over themselves Yeah - I might be a boomer.

what models are y'all using to build your products?

Who let this get to print 😂😂

Guess the market cap

@tha_kiddaa Nobody ready We’re going to Mars baby

@agenticQC Yess, and are we prepared for that? By the way I sent out a dm last night

Elon’s entering a new level we haven’t seen before

Hi Everyone: We're Vessel, the model-agnostic alternative to frontier models like ChatGPT and Claude We're built around Quality Control because we've found that people burn tokens, time, and money by using the wrong model with the wrong context Join our waitlist today

This is how we've built our deterministic QC layers Join our waitlist today to learn more and follow along our journey

This gives startups the opportunity to invest a modest amount of dollars, distill for their specific use case / problem they are solving, and then compete on price and speed It also gives them full control over the model so they are not subject to throttling / usage limits

We believe the general purpose, one-size-fits-all approach to AI will fade away: